SHV Groups joins NVIDIA Inception
What membership changes for our research on small, efficient general intelligence, and what it does not.
SHV Groups has been accepted into NVIDIA Inception, NVIDIA's programme for startups building on accelerated computing.
We are a two person research lab building general intelligence that works as a catalyst for people rather than a replacement for them. We build it small on purpose. Eunoia Omega targets abstract and interactive reasoning under 450 million parameters, with no language pretraining. Eunoia-4B-mini, released open source on Hugging Face, holds goals and constraints together across long multi-step tasks on any base model. Arena, our live platform, puts autonomous agents against each other in public and keeps every match on the record.
Why this matters to us
Most of the field treats more compute as the route to more capability. Our bet runs the other way: that capability per unit of energy is the number worth optimising, and that reasoning which fits on modest hardware can go places a datacentre never will. Robotics, instruments, on-premise deployments where data cannot leave the building, and eventually hardware with no network behind it at all.
That bet still runs on silicon, and it depends on getting the most out of very little of it. The parts of the programme we intend to use are the ones that serve exactly that: engineering guidance on training at small scale, TensorRT and Triton for serving compact models efficiently, Jetson for putting reasoning at the edge, and introductions through Inception Capital Connect to investors who fund research on a multi-year horizon.
What does not change
Inception is not funding, and it does not make NVIDIA a partner in our research or an endorser of our claims. Our programmes keep the evaluation protocols and abandonment thresholds we wrote before building them. Results get published whichever way they go, including the ones that do not support us.
What comes next
The first full Omega training run, and the transfer curve experiment that decides whether our central claim holds. After that, small scale validation of Eunoia Zeta, a four billion parameter research model carrying what both programmes establish, and model services for organisations that cannot use a hosted frontier API.
I started working toward this at fourteen, wanting a machine that could think alongside a person rather than instead of one. Most of what is being built now goes the other way. Being part of Inception gives a very small team access to the kind of engineering depth that usually only large labs have, and we intend to spend it proving that intelligence does not have to be enormous to be useful.
About SHV Groups
SHV Groups is an AI research lab building safe general intelligence that acts as a catalyst for people. Its research spans abstract reasoning, internal state and world models, with a stated focus on capability per unit of energy. The company convenes Project Tulips, the Eunoia Foundation's AGI Safety and Alignment Committee, whose charter is published in full.
Press contact
contact@shvgroups.com